A study published in *Computers in Human Behavior: Artificial Humans* found that artificial intelligence programs show high verbal reasoning abilities but struggle with visual and numerical puzzles on traditional intelligence quotient tests. Lead researcher Sherif Abdelkarim, a computer scientist at the University of California Irvine, organized the study with David Lu, Dora-Luz Flores, Susanne Jaeggi, and Pierre Baldi to measure whether advanced models possess general reasoning skills independent of specific academic knowledge. The researchers selected 18 different large language models, testing proprietary systems and open-source models. The assessment relied on a self-scoring intelligence quotient suite first published in 1996, encompassing 14 distinct categories covering three modes of thinking: verbal sections (identifying synonyms or completing complex analogies), numerical sections (solving arithmetic equations or identifying missing numbers in a sequence), and spatial rotation prompts. The results revealed wide gaps in performance depending on the format of the questions.